{"course":{"slug":"ai-for-small-business","title":"AI for a Small Business","url":"https://addaly.com/learn/ai-for-small-business","level":"beginner","track":"work"},"lesson":{"slug":"telling-customers","title":"Saying where the machine is, in one sentence","url":"https://addaly.com/learn/ai-for-small-business/telling-customers","position":84,"of":88,"minutes":9,"answer":"Disclosure is settled where a customer might otherwise believe they are talking to a person and genuinely unsettled elsewhere, so the practical rule is to disclose the interaction rather than to label every sentence.","keyIdea":"Disclosure is settled where a customer might otherwise believe they are talking to a person and genuinely unsettled elsewhere, so the practical rule is to disclose the interaction rather than to label every sentence."},"body":{"format":"markdown","text":"## Three different questions people run together\n\n**Is a customer talking to a machine right now?** This one has a clear answer. Tell them.\n\n**Was this text drafted with AI help?** Largely unsettled, and mostly nobody's business.\n\n**Was this image or voice generated?** Increasingly regulated, and moving fastest of the three.\n\nSeparating them prevents both failures: the business that discloses nothing, and the business that puts an AI notice on everything and looks either nervous or gimmicky.\n\n## Where the law is going, honestly stated\n\nThis is one of the places this course will not pretend there is a settled answer, because there is not.\n\nThe direction of travel in several jurisdictions is towards transparency obligations for systems that interact with people, and towards labelling of synthetic media. The European Union's AI Act contains transparency duties of this kind with obligations phasing in over several years; other countries have narrower rules, mostly aimed at political advertising, impersonation and sexual imagery. Some places have nothing specific at all.\n\nWhat you should take from that: the requirements differ by where your customer is, not only by where you are; they are still being interpreted; and they are tightening rather than loosening. Check what applies where you trade, and check it again next year. This is not legal advice and nothing in this course is.\n\nUnderneath the specific rules sits a much older one that already applies everywhere: consumer protection law prohibits misleading a customer about something material. A customer who believes they are speaking to a member of your staff, and is not, has been misled about something they would have wanted to know.\n\n## What to actually write\n\nA bot on your site, at the top of the conversation, in ordinary words:\n\n> You are chatting with our automated assistant. It can answer questions about orders, opening hours and stock. Type \"person\" at any time and we will get back to you within two hours.\n\nThree things doing work there. It says what it is, without a euphemism. It says what it can do, which reduces frustration more than any apology. And it gives the exit, with a real time commitment, which is the part customers actually care about.\n\nFor an automated email reply: \"This reply was generated automatically. If it has not answered your question, reply and a person will read it.\"\n\n## What not to do\n\n**Give it a human name and a photograph.** \"Hi, I'm Sarah from the support team\" is a lie, and it is the specific lie regulators are most interested in. If your assistant has a name, make it plainly a product name.\n\n**Claim to be busy, or to be checking with a colleague.** Common in bot scripts and straightforwardly deceptive.\n\n**Bury it in the terms.** Disclosure that requires reading a linked policy is not disclosure.\n\n## The part that is your choice\n\nWhether to say that a drafted email, a product description or a blog post had AI help.\n\nThere is no general obligation to, and the underlying standard is whether the output is accurate and whether you stand behind it, which you do, because you sent it. A useful test: if a customer found out later, would they feel deceived? For a routine confirmation email, no. For a handwritten-looking note of condolence, yes, and that is a piece of writing you should do yourself for reasons that precede any of this.\n\nTwo places to be more careful. Photographs of your premises, staff or products are representations about reality, and generating them crosses into misleading advertising, as the images lesson covered. And any professional body you belong to may have its own rules about work produced with AI assistance, which will bind you regardless of what general law says.\n\n## Tell your staff what to say when asked\n\nA customer asks the person on the phone whether the email they received was written by a computer. There should be an answer ready, because an improvised one usually sounds evasive.\n\nThe honest version is short: *we use software to draft some replies and a person checks every one before it is sent.* That is true, it is not an apology, and it closes the subject.\n\n## The one sentence worth having\n\nSomewhere on your site, in plain words: what you use AI for, what a person always does, and what you do with customer data.\n\nThree sentences, honestly written, and it does more for trust than any badge. Customers are not generally hostile to a business using these tools. They are hostile to finding out in a way that suggests it was being hidden."},"check":{"question":"A clinic's website assistant answers appointment questions and introduces itself as \"Priya from reception\", handing over to staff when asked. The practice manager argues no deception occurs because it hands over on request. Why is that reasoning weak?","options":[{"text":"Because patients discussing symptoms may disclose health information to the assistant believing it is covered by clinical confidentiality, which a website tool is unlikely to satisfy.","correct":false,"tempting":"An important consequence worth addressing separately, and it follows from the false impression rather than explaining why the impression is the fault."},{"text":"Because the handover depends on the patient thinking to ask, and a patient who believes Priya is a receptionist has no reason to ask for a person.","correct":false,"tempting":"Very close, and it explains why the exit fails rather than why the introduction is already a misrepresentation."},{"text":"Because using a human name creates an expectation of accountability, and patients cannot pursue a complaint about advice given by someone who does not exist.","correct":false,"tempting":"A real problem when something goes wrong, and it describes a downstream difficulty rather than the misrepresentation itself."},{"text":"Because the name and role already assert a person exists, and that has happened before any handover is offered.","correct":true}]},"prev":{"slug":"when-somebody-leaves","title":"The week somebody leaves","url":"https://addaly.com/learn/ai-for-small-business/when-somebody-leaves"},"next":{"slug":"what-you-warrant","title":"What you have already promised your clients","url":"https://addaly.com/learn/ai-for-small-business/what-you-warrant"},"authorship":"Founder and series editor: Dr Mustafa Mun. Written by AI models to a written standard. Exam keys checked blind by a second model: 3,665 of 3,666 agreed. Academic review invited.","review":{"status":"Draft, open for academic review"},"cite":"Addaly (2026) 'Saying where the machine is, in one sentence', in AI for a Small Business. Edited by M. Mun. Addaly. Available at: https://addaly.com/learn/ai-for-small-business/telling-customers (Accessed: 13 October 2026).","licence":{"name":"CC BY-NC-SA 4.0","url":"https://creativecommons.org/licenses/by-nc-sa/4.0/"},"reader":{"schema":1,"title":"Saying where the machine is, in one sentence","courseTitle":"AI for a Small Business","module":{"index":9,"slug":"cost-risk-and-the-long-run","title":"What it costs, what it risks, and what happens in a year"},"answer":null,"keyIdea":"Disclosure is settled where a customer might otherwise believe they are talking to a person and genuinely unsettled elsewhere, so the practical rule is to disclose the interaction rather than to label every sentence.","hero":{"kind":"plate","image":null,"alt":null,"caption":null},"blocks":[{"t":"h2","id":"three-different-questions-people-run-together","runs":[{"text":"Three different questions people run together"}]},{"t":"p","runs":[{"text":"Is a customer talking to a machine right now?","b":true},{"text":" This one has a clear answer. Tell them."}]},{"t":"p","runs":[{"text":"Was this text drafted with AI help?","b":true},{"text":" Largely unsettled, and mostly nobody’s business."}]},{"t":"p","runs":[{"text":"Was this image or voice generated?","b":true},{"text":" Increasingly regulated, and moving fastest of the three."}]},{"t":"p","runs":[{"text":"Separating them prevents both failures: the business that discloses nothing, and the business that puts an AI notice on everything and looks either nervous or gimmicky."}]},{"t":"h2","id":"where-the-law-is-going-honestly-stated","runs":[{"text":"Where the law is going, honestly stated"}]},{"t":"p","runs":[{"text":"This is one of the places this course will not pretend there is a settled answer, because there is not."}]},{"t":"p","runs":[{"text":"The direction of travel in several jurisdictions is towards transparency obligations for systems that interact with people, and towards labelling of synthetic media. The European Union’s AI Act contains transparency duties of this kind with obligations phasing in over several years; other countries have narrower rules, mostly aimed at political advertising, impersonation and sexual imagery. Some places have nothing specific at all."}]},{"t":"p","runs":[{"text":"What you should take from that: the requirements differ by where your customer is, not only by where you are; they are still being interpreted; and they are tightening rather than loosening. Check what applies where you trade, and check it again next year. This is not legal advice and nothing in this course is."}]},{"t":"p","runs":[{"text":"Underneath the specific rules sits a much older one that already applies everywhere: consumer protection law prohibits misleading a customer about something material. A customer who believes they are speaking to a member of your staff, and is not, has been misled about something they would have wanted to know."}]},{"t":"h2","id":"what-to-actually-write","runs":[{"text":"What to actually write"}]},{"t":"p","runs":[{"text":"A bot on your site, at the top of the conversation, in ordinary words:"}]},{"t":"quote","runs":[{"text":"You are chatting with our automated assistant. It can answer questions about orders, opening hours and stock. Type “person” at any time and we will get back to you within two hours."}]},{"t":"p","runs":[{"text":"Three things doing work there. It says what it is, without a euphemism. It says what it can do, which reduces frustration more than any apology. And it gives the exit, with a real time commitment, which is the part customers actually care about."}]},{"t":"p","runs":[{"text":"For an automated email reply: “This reply was generated automatically. If it has not answered your question, reply and a person will read it.”"}]},{"t":"h2","id":"what-not-to-do","runs":[{"text":"What not to do"}]},{"t":"p","runs":[{"text":"Give it a human name and a photograph.","b":true},{"text":" “Hi, I’m Sarah from the support team” is a lie, and it is the specific lie regulators are most interested in. If your assistant has a name, make it plainly a product name."}]},{"t":"p","runs":[{"text":"Claim to be busy, or to be checking with a colleague.","b":true},{"text":" Common in bot scripts and straightforwardly deceptive."}]},{"t":"p","runs":[{"text":"Bury it in the terms.","b":true},{"text":" Disclosure that requires reading a linked policy is not disclosure."}]},{"t":"h2","id":"the-part-that-is-your-choice","runs":[{"text":"The part that is your choice"}]},{"t":"p","runs":[{"text":"Whether to say that a drafted email, a product description or a blog post had AI help."}]},{"t":"p","runs":[{"text":"There is no general obligation to, and the underlying standard is whether the output is accurate and whether you stand behind it, which you do, because you sent it. A useful test: if a customer found out later, would they feel deceived? For a routine confirmation email, no. For a handwritten-looking note of condolence, yes, and that is a piece of writing you should do yourself for reasons that precede any of this."}]},{"t":"p","runs":[{"text":"Two places to be more careful. Photographs of your premises, staff or products are representations about reality, and generating them crosses into misleading advertising, as the images lesson covered. And any professional body you belong to may have its own rules about work produced with AI assistance, which will bind you regardless of what general law says."}]},{"t":"h2","id":"tell-your-staff-what-to-say-when-asked","runs":[{"text":"Tell your staff what to say when asked"}]},{"t":"p","runs":[{"text":"A customer asks the person on the phone whether the email they received was written by a computer. There should be an answer ready, because an improvised one usually sounds evasive."}]},{"t":"p","runs":[{"text":"The honest version is short: "},{"text":"we use software to draft some replies and a person checks every one before it is sent.","i":true},{"text":" That is true, it is not an apology, and it closes the subject."}]},{"t":"h2","id":"the-one-sentence-worth-having","runs":[{"text":"The one sentence worth having"}]},{"t":"p","runs":[{"text":"Somewhere on your site, in plain words: what you use AI for, what a person always does, and what you do with customer data."}]},{"t":"p","runs":[{"text":"Three sentences, honestly written, and it does more for trust than any badge. Customers are not generally hostile to a business using these tools. They are hostile to finding out in a way that suggests it was being hidden."}]}],"check":{"question":[{"text":"A clinic’s website assistant answers appointment questions and introduces itself as “Priya from reception”, handing over to staff when asked. The practice manager argues no deception occurs because it hands over on request. Why is that reasoning weak?"}],"options":[{"text":[{"text":"Because patients discussing symptoms may disclose health information to the assistant believing it is covered by clinical confidentiality, which a website tool is unlikely to satisfy."}],"correct":false,"tempting":[{"text":"An important consequence worth addressing separately, and it follows from the false impression rather than explaining why the impression is the fault."}]},{"text":[{"text":"Because the handover depends on the patient thinking to ask, and a patient who believes Priya is a receptionist has no reason to ask for a person."}],"correct":false,"tempting":[{"text":"Very close, and it explains why the exit fails rather than why the introduction is already a misrepresentation."}]},{"text":[{"text":"Because using a human name creates an expectation of accountability, and patients cannot pursue a complaint about advice given by someone who does not exist."}],"correct":false,"tempting":[{"text":"A real problem when something goes wrong, and it describes a downstream difficulty rather than the misrepresentation itself."}]},{"text":[{"text":"Because the name and role already assert a person exists, and that has happened before any handover is offered."}],"correct":true,"tempting":null}],"order":[1,2,3,0],"why":null},"prev":{"slug":"when-somebody-leaves","title":"The week somebody leaves","minutes":8},"next":{"slug":"what-you-warrant","title":"What you have already promised your clients","minutes":8}}}